CN115693918B

CN115693918B - Comprehensive intelligent electricity utilization system and method for building

Info

Publication number: CN115693918B
Application number: CN202211099055.3A
Authority: CN
Inventors: 许建成; 唐平亚
Original assignee: Zhejiang Xinyou Electromechanical Equipment Installation Co ltd
Current assignee: Zhejiang Xinyou Electromechanical Equipment Installation Co ltd
Priority date: 2022-09-07
Filing date: 2022-09-07
Publication date: 2023-08-18
Anticipated expiration: 2042-09-07
Also published as: CN115693918A

Abstract

The method comprises the steps of taking a first convolutional neural network model as a feature extractor to extract implicit association features of state information of all electric equipment in a building to be monitored at a current time point, extracting global semantic association features described by type texts of all electric equipment in all electric equipment through a context encoder comprising an embedded layer, then fusing the features through a graph neural network, and finally, passing a state semantic association feature matrix obtained after fusion through a classifier to obtain a classification result for indicating whether state combinations of all the electric equipment in the building to be monitored are normal or not, and generating an abnormal power utilization prompt for the state combinations of all the electric equipment in the building to be monitored according to the classification result. Therefore, the abnormality of the state combination of all the electric equipment in the building can be accurately analyzed and judged, and then an early warning prompt is generated for the abnormal situation.

Description

Comprehensive intelligent electricity utilization system and method for building

Technical Field

The application relates to the technical field of smart power grids, in particular to a comprehensive intelligent electricity utilization system for a building and a method thereof.

Background

Due to rapid development of social economy, people have higher and higher requirements on the comfort of the building environment, so that building energy consumption is higher and higher, and how to improve the comfort of the building environment without improving the building energy consumption is a research target of green building, and even the energy-saving effect can be achieved.

The intelligent electricity utilization technology is an important component for building a strong intelligent power grid, and the intelligent electricity utilization technology is characterized in that intelligent service is realized, diversified requirements of users are met, stable, reliable, economical and safe power supply is realized, and a novel electricity supply and utilization relation of real-time interaction of power flow, information flow and business flow between the power grid and the clients is constructed. The energy utilization mode of the user is changed, the energy conservation and emission reduction are promoted, and the specific gravity of clean electric energy in terminal energy consumption is improved. At present, the electricity utilization technology of China is relatively backward, most of the electricity utilization technology adopts the traditional switch, and the intelligent electricity utilization technology is less in application, so that the electricity utilization technology is an important aspect of energy consumption waste.

The emerging Internet of things technology can organically combine multi-energy optimization with intelligent electricity consumption, so that the purposes of scientific electricity consumption and open source throttling are achieved, and the Internet of things technology is widely applied to buildings, so that the intelligent and energy-saving trend of the buildings is achieved.

Therefore, an optimized building integrated intelligent electricity usage scheme is desired.

Disclosure of Invention

The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a comprehensive intelligent electricity utilization system of a building and a method thereof, wherein a first convolutional neural network model is used as a feature extractor to extract implicit association features of state information of all electric equipment in the building to be monitored at a current time point, a context encoder comprising an embedded layer is used to extract global semantic association features of type text description of each electric equipment in all electric equipment, then a graph neural network is used to fuse the features, finally a state semantic association feature matrix obtained after the fusion is used to obtain a classification result used for indicating whether the state combination of all electric equipment in the building to be monitored is normal or not, and an abnormal electricity utilization prompt is generated for the state combination of all electric equipment in the building to be monitored in response to the classification result. Therefore, the abnormality of the state combination of all the electric equipment in the building can be accurately analyzed and judged, and then an early warning prompt is generated for the abnormal situation.

According to one aspect of the present application, there is provided a building integrated intelligent electricity system, comprising:

The state information acquisition unit is used for acquiring state information of all electric equipment in the building to be monitored at the current time point;

the state association unit is used for arranging the state information of all electric equipment in the building to be monitored at the current time point into a state input vector, and calculating the product between the state input vector and the transposed vector thereof to obtain a state association matrix;

the state association feature extraction unit is used for obtaining a state association feature matrix through a first convolutional neural network model serving as a feature extractor, wherein the first convolutional neural network model is trained by the state association matrix;

the device description information acquisition unit is used for acquiring the type text description of each electric equipment in all the electric equipment;

the device description semantic coding unit is used for respectively obtaining a plurality of device text description feature vectors through the trained context encoder comprising the embedded layer by the type text description of each electric device in all electric devices;

a matrix construction unit, configured to arrange the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension;

the graphic neural network unit is used for training the equipment text description feature matrix and the state association feature matrix to obtain a state semantic association feature matrix through a graphic neural network, wherein the graphic neural network encodes the equipment text description feature matrix and the state association feature matrix through a learnable neural network parameter to obtain equipment text description association feature representation containing irregular state topology information;

The power consumption equipment state monitoring result generating unit is used for obtaining a classification result by training the state semantic association feature matrix through a classifier, wherein the classification result is used for indicating whether the state combination of all power consumption equipment in a building to be monitored is normal or not; and

and the electricity utilization control result generating unit is used for responding to the classification result that the state combination of all electric equipment in the building to be monitored is abnormal and generating an electricity utilization abnormality prompt.

In the above building integrated intelligent electricity utilization system, the state association feature extraction unit is further configured to: each layer of the trained first convolutional neural network model serving as the feature extractor is respectively carried out in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; carrying out mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the first convolutional neural network model serving as the feature extractor after training is the state association feature matrix, and the input of the first layer of the first convolutional neural network model serving as the feature extractor after training is the state association matrix.

In the above building integrated intelligent electricity utilization system, the device description semantic coding unit includes:

the word segmentation subunit is used for carrying out word segmentation processing on the type text description of each electric equipment in all electric equipment so as to convert the type text description of each electric equipment in all electric equipment into a word sequence consisting of a plurality of words; an embedded encoding subunit for mapping each word in the word sequence to a word vector using an embedding layer of the context encoder to obtain a sequence of word vectors; and a context encoding subunit configured to perform global-based context semantic encoding on the sequence of word vectors using a translator of the context encoder to obtain the plurality of device text description feature vectors.

In the above building integrated intelligent power consumption system, the power consumption device state monitoring result generating unit is further configured to: processing the state semantic association feature matrix using the classifier to generate the classification result with the following formula:wherein->Representing projection of the state semantic association feature matrix as a vector,>weight matrix for full connection layer, +. >Representing the bias matrix of the fully connected layer.

In the above building integrated intelligent power consumption system, the building integrated intelligent power consumption system further comprises a training module for performing joint training on the first convolutional neural network model as the feature extractor, the context encoder including the embedded layer, the graph neural network and the classifier.

In the above-mentioned building integrated intelligent power consumption system, the training module includes: the training data acquisition unit is used for acquiring training data, wherein the training data comprises state information of all electric equipment in the building to be monitored at a preset time point and type text description of each electric equipment in the electric equipment; the training state association unit is used for arranging the state information of all electric equipment in the building to be monitored in the training data at a preset time point into a training state input vector, and calculating the product between the training state input vector and the transposed vector thereof to obtain a training state association matrix; the training state association feature extraction unit is used for enabling the training state association matrix to pass through the first convolution neural network model serving as the feature extractor to obtain a training state association feature matrix; the training equipment description semantic coding unit is used for respectively obtaining a plurality of training equipment text description feature vectors through the context encoder comprising the embedded layer by using the type text description of each electric equipment in all electric equipment in the training data; the training matrix construction unit is used for arranging the training device text description feature vectors into a training device text description feature matrix according to the device sample dimension; the training image neural network unit is used for enabling the training equipment text description feature matrix and the training state association feature matrix to pass through the image neural network so as to obtain a training state semantic association feature matrix; the matrix unfolding unit is used for unfolding the training state semantic association feature matrix into classification feature vectors according to row vectors; the classification vector correction unit is used for correcting the classification feature vector by using the weight matrix before and after each iteration update of the classifier so as to obtain a corrected classification feature vector; the classification loss unit is used for passing the corrected classification characteristic vector through the classifier to obtain a classification loss function value; and a parameter updating unit for jointly training the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network and the classifier with the classification loss function value as a loss function value and through back propagation of gradient descent.

In the above building integrated intelligent electricity utilization system, the classification vector correction unit is further configured to: correcting the classification characteristic vector by using a weight matrix of the classifier before and after each iteration update according to the following formula to obtain the corrected classification characteristic vector; wherein, the formula is:

wherein the method comprises the steps ofRepresenting the classification feature vector,/->And->Respectively representing the weight matrix before and after each iteration update of the classifier, and +.>Representing zero norm of vector,/->Representing addition by position->Representing difference by position +.>Representing matrix multiplication +.>An exponential operation representing a vector that represents a calculation of a natural exponential function value that is a power of a eigenvalue of each position in the vector.

According to another aspect of the present application, there is provided a building integrated intelligent electricity using method, comprising:

acquiring state information of all electric equipment in a building to be monitored at a current time point;

after state information of all electric equipment in the building to be monitored at the current time point is arranged into a state input vector, calculating the product between the state input vector and the transposed vector of the state input vector to obtain a state association matrix;

The state association matrix is obtained through a first convolution neural network model which is completed by training and serves as a feature extractor;

acquiring the type text description of each electric equipment in all electric equipment;

the text description of the type of each electric equipment in all electric equipment is respectively passed through a context encoder which is completed by training and comprises an embedded layer so as to obtain a plurality of equipment text description feature vectors;

arranging the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension;

training the device text description feature matrix and the state association feature matrix through a trained graphic neural network to obtain a state semantic association feature matrix, wherein the graphic neural network encodes the device text description feature matrix and the state association feature matrix through a learnable neural network parameter to obtain a device text description association feature representation containing irregular state topology information;

the state semantic association feature matrix is subjected to training to obtain a classification result by a classifier, wherein the classification result is used for indicating whether the state combination of all electric equipment in a building to be monitored is normal or not; and

And generating an abnormal electricity utilization prompt in response to the classification result that the state combination of all electric equipment in the building to be monitored is abnormal.

Compared with the prior art, the comprehensive intelligent electricity utilization system and the comprehensive intelligent electricity utilization method for the building are characterized in that the first convolutional neural network model is used as a feature extractor to extract implicit association features of state information of all electric equipment in a building to be monitored at a current time point, a context encoder comprising an embedded layer is used for extracting global semantic association features of type text description of each electric equipment in all electric equipment, then the graphic neural network is used for feature fusion, finally the state semantic association feature matrix obtained after fusion is used for obtaining a classification result used for indicating whether state combinations of all electric equipment in the building to be monitored are normal or not through a classifier, and abnormal electricity utilization prompt is generated for the state combinations of all electric equipment in the building to be monitored in response to the classification result. Therefore, the abnormality of the state combination of all the electric equipment in the building can be accurately analyzed and judged, and then an early warning prompt is generated for the abnormal situation.

Drawings

The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.

FIG. 1 illustrates an application scenario diagram of a building integrated intelligent electricity system according to an embodiment of the present application.

FIG. 2 illustrates a block diagram of a building integrated intelligent electricity system, according to an embodiment of the present application.

FIG. 3 illustrates a block diagram of the device description semantic coding unit in the building integrated intelligent electricity usage system according to an embodiment of the present application.

FIG. 4 illustrates a block diagram of the training module in the building integrated intelligent power consumption system, according to an embodiment of the application.

Fig. 5 illustrates a flow chart of a building integrated intelligent electricity usage method according to an embodiment of the present application.

Fig. 6 illustrates an architectural diagram of a building integrated intelligent electricity usage method according to an embodiment of the present application.

Detailed Description

Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

Scene overview

Accordingly, when the technology of the internet of things is utilized to organically combine multi-energy optimization and intelligent electricity utilization in a building, the condition combination of all electric equipment in the building is required to be subjected to anomaly analysis and detection. However, at present, there is a hidden state association between some electric devices in a building, for example, a curtain and a lamp are generally turned on and off simultaneously, so that the hidden state association between the electric devices needs to be fully mined to determine the state of the electric devices in the building, and then a prompt signal is generated when the state is determined to be abnormal. Therefore, after receiving the prompt signal, the user can adjust the state of the corresponding electric equipment so as to realize energy saving.

Based on the above, in the technical scheme of the application, the state implicit association features of all the electric equipment in the building and the global semantic association features of the type text description of each electric equipment in all the electric equipment are respectively extracted by using a deep neural network model, and the fusion of the features is further performed by using a graph neural network so as to strengthen and optimize the classification effect of the state association features of each electric equipment based on the global type text description features of each electric equipment, thereby improving the accuracy of detecting and judging the state abnormality of all the electric equipment in the building.

Specifically, in the technical scheme of the application, firstly, the state information of all electric equipment in a building to be monitored at the current time point is obtained. Here, if the electric equipment is in the on state, the value of the corresponding position is 1, and if the electric equipment is in the off state, the value of the corresponding position is 0.

And then, arranging the state information of all the electric equipment in the building to be monitored at the current time point into a state input vector. It should be understood that, in consideration of the fact that there may be implicit correlation features between states of each electric device in the building, for example, the window curtain and the lamp may be turned on and off at the same time, so as to fully mine such implicit correlation during feature mining to determine an abnormal state of each electric device in the building, further calculate a product between the state input vector and the transpose vector thereof to obtain a state correlation matrix, and further construct state correlation information between each electric device. And then, carrying out deep association feature mining on the state association matrix by using a first convolutional neural network model which is used as a feature extractor and has excellent performance in the aspect of implicit feature extraction so as to extract the implicit association features of all electric equipment in the building to be monitored, thereby obtaining the state association feature matrix.

Further, considering that each of all the electric devices in the building has different types, and using type features to strengthen relevance features of each electric device obviously can improve classification accuracy. Therefore, in the technical scheme of the application, the semantic features of the type text description of each electric equipment in all electric equipment are selected to strengthen and optimize the state relevance of each electric equipment. The method comprises the steps of obtaining type text descriptions of all electric equipment, coding the type text descriptions of all electric equipment in all electric equipment through a context coder comprising an embedded layer, extracting global-based high-dimensional semantic features of the types of all electric equipment in all electric equipment to be more suitable for representing essential type relevance features among the electric equipment, and obtaining a plurality of equipment text description feature vectors. And then arranging the plurality of device text description feature vectors corresponding to the electric devices into a device text description feature matrix according to the device sample dimension so as to integrate the relevance features of the electric devices based on the global electric device types.

It should be appreciated that each row vector in the device text description feature matrix corresponds to a type feature of each powered device (node), and the state association feature matrix represents a high-dimensional implicit feature representation of a state association (i.e., an association between nodes) between respective two of the powered devices. Therefore, in the technical scheme of the application, the equipment text description feature matrix and the state association feature matrix construct a graph data structure. Therefore, in the technical scheme of the application, the device text description feature matrix and the state association feature matrix are encoded by using the graph neural network model as a feature encoder through the learnable neural network parameters so as to obtain the device text description association feature representation containing irregular state topology information.

And then, the state semantic association feature matrix is trained by the classifier to obtain a classification result for indicating whether the state combination of all electric equipment in the building to be monitored is normal. Correspondingly, in a specific example of the application, in response to the classification result being that the state combination of all electric equipment in the building to be monitored is abnormal, an abnormal electricity utilization prompt is generated.

In particular, in the technical scheme of the application, for the state semantic association feature matrix obtained by passing the device text description feature matrix and the state association feature matrix through the graph neural network, as the state semantic association feature vector of the state semantic association feature matrix perpendicular to the sample dimension represents the state topology text description feature of a single electric device, the correlation of the state topology text description feature among a plurality of devices is possibly lower, thereby increasing the difficulty of adapting the parameters of the weight matrix of the classifier and the classification feature vector in the training process of the classifier.

Based on this, in the technical solution of the present application, the parameters of the weight matrix of the classifier are adjusted in the training process, and the feature vector obtained after the state semantic association feature matrix is expanded is, for example, written asScene-dependent optimization of classifier iterations is performed, expressed as:

and->Is the weight matrix of the classifier before and after each iteration update, < >>Representing the zero norm of the vector.

Here, the iterative scene correlation optimization of the classifier uses the measure of the scene point correlation before and after the parameter update of the weight matrix of the classifier at the time of iteration as a correction factor to optimize the class probability representation of the classification feature vector, so as to make a support for the correlation description of the classification feature vector through the distribution similarity of the classification scene of the classifier, and to promote the suitability between the parameter of the weight matrix of the classifier and the classification feature vector from the angle of the classification feature vector, thus, the training speed of the classifier and the accuracy of the classification result of the classification feature vector can be promoted by simultaneously regulating the parameter of the weight matrix of the classifier and the parameter of the classification feature vector. Therefore, the abnormality of the state combination of all the electric equipment in the building can be analyzed and judged, and then an early warning prompt is generated for the abnormal situation, so that after receiving the prompt signal, a user can adjust the state of the corresponding electric equipment to realize energy saving.

Based on this, the application provides a comprehensive intelligent electricity utilization system for buildings, which comprises: the state information acquisition unit is used for acquiring state information of all electric equipment in the building to be monitored at the current time point;

the state association unit is used for arranging the state information of all electric equipment in the building to be monitored at the current time point into a state input vector, and calculating the product between the state input vector and the transposed vector thereof to obtain a state association matrix; the state association feature extraction unit is used for obtaining a state association feature matrix through a first convolutional neural network model serving as a feature extractor, wherein the first convolutional neural network model is trained by the state association matrix; the device description information acquisition unit is used for acquiring the type text description of each electric equipment in all the electric equipment; the device description semantic coding unit is used for respectively obtaining a plurality of device text description feature vectors through the trained context encoder comprising the embedded layer by the type text description of each electric device in all electric devices; a matrix construction unit, configured to arrange the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension;

The graphic neural network unit is used for training the equipment text description feature matrix and the state association feature matrix to obtain a state semantic association feature matrix through a graphic neural network, wherein the graphic neural network encodes the equipment text description feature matrix and the state association feature matrix through a learnable neural network parameter to obtain equipment text description association feature representation containing irregular state topology information; the power consumption equipment state monitoring result generating unit is used for obtaining a classification result by training the state semantic association feature matrix through a classifier, wherein the classification result is used for indicating whether the state combination of all power consumption equipment in a building to be monitored is normal or not; and the electricity utilization control result generating unit is used for responding to the classification result that the state combination of all the electric equipment in the building to be monitored is abnormal and generating an electricity utilization abnormality prompt.

FIG. 1 illustrates an application scenario diagram of a building integrated intelligent electricity system according to an embodiment of the present application. As shown in fig. 1, in the application scenario, state information (e.g., M as illustrated in fig. 1) of all electric devices (e.g., T as illustrated in fig. 1) in a building to be monitored at a current time point is acquired, and a type text description (e.g., C as illustrated in fig. 1) of each of all electric devices; and then inputting the acquired state information of all the electric equipment at the current time point and the type text description of each electric equipment into a server (for example, S as shown in fig. 1) deployed with a building comprehensive intelligent electricity utilization system, wherein the server processes the state information of all the electric equipment at the current time point and the type text description of each electric equipment by using a building comprehensive intelligent electricity utilization algorithm to generate a classification result for indicating whether the state combination of all the electric equipment in a building to be monitored is normal or not, and generating an electricity utilization abnormality prompt in response to the classification result that the state combination of all the electric equipment in the building to be monitored is abnormal.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Exemplary System

FIG. 2 illustrates a block diagram of a building integrated intelligent electricity system, according to an embodiment of the present application. As shown in fig. 2, the building integrated intelligent electricity system 100 according to the embodiment of the present application includes: the state information acquisition unit 110 is configured to acquire state information of all electric devices in the building to be monitored at a current time point; the state association unit 120 is configured to arrange state information of all electric devices in the building to be monitored at a current time point into a state input vector, and calculate a product between the state input vector and a transpose vector thereof to obtain a state association matrix; a state association feature extraction unit 130, configured to obtain a state association feature matrix by using the state association matrix through a first convolutional neural network model that is a feature extractor and is completed through training; the device description information collection unit 140 is configured to obtain a type text description of each of the all electric devices; the device description semantic coding unit 150 is configured to obtain a plurality of device text description feature vectors by respectively training a context encoder including an embedded layer, where the context encoder includes an embedded layer, for each type text description of each of the all electric devices; a matrix construction unit 160, configured to arrange the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension; a graphic neural network unit 170, configured to train the device text description feature matrix and the state association feature matrix to obtain a state semantic association feature matrix through a graphic neural network, where the graphic neural network encodes the device text description feature matrix and the state association feature matrix through a learnable neural network parameter to obtain a device text description association feature representation including irregular state topology information; the electric equipment state monitoring result generating unit 180 is configured to obtain a classification result by training the state semantic association feature matrix through a classifier, where the classification result is used to indicate whether the state combinations of all electric equipment in the building to be monitored are normal; and a power utilization control result generating unit 190, configured to generate a power utilization abnormality prompt in response to the classification result being that the state combination of all the electric devices in the building to be monitored is abnormal.

Specifically, in the embodiment of the present application, the status information collection unit 110 is configured to obtain status information of all electric devices in the building to be monitored at a current time point. As can be seen from the foregoing, it is considered that when the internet of things technology is used to organically combine multi-energy optimization and intelligent electricity utilization in a building, it is necessary to perform anomaly analysis and detection on the state combinations of all the electric devices in the building. However, at present, there is a hidden state association between some electric devices in a building, for example, a curtain and a lamp are generally turned on and off simultaneously, so that the hidden state association between the electric devices needs to be fully mined to determine the state of the electric devices in the building, and then a prompt signal is generated when the state is determined to be abnormal. Therefore, after receiving the prompt signal, the user can adjust the state of the corresponding electric equipment so as to realize energy saving.

Specifically, in the embodiment of the present application, the state association unit 120 is configured to, after arranging the state information of all the electric devices in the building to be monitored at the current time point into a state input vector, calculate a product between the state input vector and a transposed vector thereof to obtain a state association matrix. That is, after the state information of all the electric devices in the building to be monitored at the current time point is obtained, the state information of all the electric devices in the building to be monitored at the current time point is arranged into a state input vector. It should be appreciated that by constructing the state information of all the electric devices in the building to be monitored at the current time point into a vector form, the subsequent computer processing is facilitated.

It should be understood that, in consideration of the fact that there may be implicit correlation features between states of each electric device in the building, for example, the window curtain and the lamp may be turned on and off at the same time, so as to fully mine such implicit correlation during feature mining to determine an abnormal state of each electric device in the building, further calculate a product between the state input vector and the transpose vector thereof to obtain a state correlation matrix, and further construct state correlation information between each electric device.

Specifically, in the embodiment of the present application, the state-associated feature extraction unit 130 is configured to obtain the state-associated feature matrix by using the trained first convolutional neural network model serving as the feature extractor for the state-associated feature matrix. The first convolutional neural network model which is used as a feature extractor and has excellent performance in the aspect of implicit feature extraction is used for carrying out deep associated feature mining on the state associated matrix so as to extract the implicit associated features of all electric equipment in the building to be monitored, and thus the state associated feature matrix is obtained.

More specifically, in the embodiment of the present application, each layer of the trained first convolutional neural network model serving as the feature extractor is performed in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; carrying out mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the first convolutional neural network model serving as the feature extractor after training is the state association feature matrix, and the input of the first layer of the first convolutional neural network model serving as the feature extractor after training is the state association matrix.

That is, the state correlation matrix is input to the first layer of the trained first convolutional neural network model as a feature extractor first, then each layer of the trained first convolutional neural network model as a feature extractor performs convolution processing, averaging processing and nonlinear activation processing on input data in forward transfer of the layer, and finally the output of the last layer of the trained first convolutional neural network model as a feature extractor is the state correlation feature matrix.

Specifically, in the embodiment of the present application, the device description information collection unit 140 and the device description semantic coding unit 150 are configured to obtain the type text descriptions of each of the all the electric devices, and are configured to obtain a plurality of device text description feature vectors by respectively training the context encoders including the embedded layer, where the context encoders are completed for each of the all the electric devices. It should be appreciated that considering that each of all the consumers within the building are of a different type, and using type features to enhance the relevance features of each consumer obviously can improve the accuracy of classification. Therefore, in the technical scheme of the application, the semantic features of the type text description of each electric equipment in all electric equipment are selected to strengthen and optimize the state relevance of each electric equipment.

In the technical scheme of the application, the type text description of each electric equipment in all electric equipment is obtained, and the type text description of each electric equipment in all electric equipment is encoded in a context encoder comprising an embedded layer, so that the global-based high-dimensional semantic features of the types of each electric equipment in all electric equipment are extracted to be more suitable for representing the essential type relevance features among the electric equipment, and a plurality of equipment text description feature vectors are obtained.

More specifically, in an embodiment of the present application, fig. 3 illustrates a block diagram of the device description semantic coding unit in the building integrated intelligent electricity using system according to an embodiment of the present application, and as shown in fig. 3, the device description semantic coding unit 150 includes: a word segmentation subunit 210, configured to perform word segmentation processing on the type text description of each of the all electric devices to convert the type text description of each of the all electric devices into a word sequence composed of a plurality of words; an embedded encoding subunit 220, configured to map each word in the word sequence to a word vector using an embedding layer of the context encoder to obtain a sequence of word vectors; and a context encoding subunit 230 configured to perform global-based context semantic encoding on the sequence of word vectors using a translator of the context encoder to obtain the plurality of device text description feature vectors.

In particular, the word segmentation refers to a process of segmenting a Chinese character sequence into individual words, that is, recombining continuous word sequences into word sequences according to a certain specification. Specifically, in a specific example of the present application, an understanding-based word segmentation method may be selected, where the understanding-based word segmentation method achieves the effect of recognizing words by letting a computer simulate understanding of sentences by a person. In another specific example of the present application, a word segmentation method based on statistics may also be selected, where the word segmentation method based on statistics is to learn a word segmentation rule by using a statistical machine learning model on the premise of giving a large number of segmented texts, so as to achieve segmentation of unknown texts.

It should be understood that, in the technical solution of the present application, the context encoder uses a Bert model based on a converter to perform global context semantic encoding on the sequence of word vectors, and in particular, the Bert model uses an intrinsic mask structure of a converter to perform global context semantic encoding on each word vector in the sequence of word vectors with global context of the sequence of word vectors as a semantic context to obtain the plurality of device text description feature vectors.

Specifically, in the embodiment of the present application, the matrix construction unit 160 is configured to arrange the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension. That is, after the device text description feature vectors are obtained, the device text description feature vectors corresponding to the electric devices are arranged into a device text description feature matrix according to the device sample dimension, so as to integrate relevance features of the electric devices based on the global electric device types.

Specifically, in the embodiment of the present application, the graphic neural network unit 170 is configured to train the device text description feature matrix and the state association feature matrix to obtain a state semantic association feature matrix through a graphic neural network completed by training, where the graphic neural network encodes the device text description feature matrix and the state association feature matrix through a learnable neural network parameter to obtain a device text description association feature representation including irregular state topology information. It should be appreciated that each row vector in the device text description feature matrix corresponds to a type feature of each powered device (node), and the state association feature matrix represents a high-dimensional implicit feature representation of a state association (i.e., an association between nodes) between respective two of the powered devices.

Therefore, in the technical scheme of the application, the equipment text description feature matrix and the state association feature matrix construct a graph data structure. Specifically, the device text description feature matrix and the state association feature matrix are encoded by a learnable neural network parameter using a graph neural network model as a feature encoder to obtain a device text description association feature representation containing irregular state topology information.

Specifically, in the embodiment of the present application, the power consumption device state monitoring result generating unit 180 and the power consumption control result generating unit 190 are configured to obtain a classification result by training the state semantic association feature matrix through a classifier completed, where the classification result is used to indicate whether the state combinations of all power consumption devices in the building to be monitored are normal, and generate a power consumption abnormality prompt in response to the classification result that the state combinations of all power consumption devices in the building to be monitored are abnormal.

The state semantic association feature matrix is trained to obtain a classification result used for indicating whether the state combination of all electric equipment in the building to be monitored is normal or not. Correspondingly, in a specific example of the application, in response to the classification result being that the state combination of all electric equipment in the building to be monitored is abnormal, an abnormal electricity utilization prompt is generated.

More specifically, in the embodiment of the present application, the classifier is used to process the state semantic association feature matrix according to the following formula to generate the classification result, where the formula is:wherein->Representing the shape of the objectProjection of a state semantic association feature matrix as a vector, +.>Weight matrix for full connection layer, +.>Representing the bias matrix of the fully connected layer.

Further, in the technical solution of the present application, the building integrated intelligent power consumption system further includes a training module 300 for performing joint training on the first convolutional neural network model as the feature extractor, the context encoder including the embedded layer, the graph neural network, and the classifier.

Wherein fig. 4 illustrates a block diagram of the training module in the building integrated intelligent electricity system according to an embodiment of the present application, as shown in fig. 4, the training module 300 includes: the training data acquisition unit 301 is configured to acquire training data, where the training data includes status information of all electric devices in the building to be monitored at a predetermined time point and a type text description of each electric device in the all electric devices; the training state association unit 302 is configured to arrange state information of all electric devices in the building to be monitored in the training data at a predetermined time point into a training state input vector, and calculate a product between the training state input vector and a transpose vector thereof to obtain a training state association matrix; a training state association feature extraction unit 303, configured to pass the training state association matrix through the first convolutional neural network model serving as a feature extractor to obtain a training state association feature matrix; the training device description semantic coding unit 304 is configured to obtain a plurality of training device text description feature vectors by respectively passing the type text descriptions of each of the all the electric devices in the training data through the context encoder including the embedded layer; a training matrix construction unit 305, configured to arrange the plurality of training device text description feature vectors into a training device text description feature matrix according to the device sample dimension; a training graph neural network unit 306, configured to pass the training device text description feature matrix and the training state association feature matrix through the graph neural network to obtain a training state semantic association feature matrix; a matrix expansion unit 307, configured to expand the training state semantic association feature matrix into a classification feature vector according to a row vector; a classification vector correction unit 308, configured to correct the classification feature vector by using the weight matrix before and after each iteration update of the classifier to obtain a corrected classification feature vector; a classification loss unit 309, configured to pass the corrected classification feature vector through the classifier to obtain a classification loss function value; and a parameter updating unit 310, configured to jointly train the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network, and the classifier with the classification loss function value as a loss function value and through back propagation of gradient descent.

Particularly, in the technical scheme of the application, for the state semantic association feature matrix obtained by passing the device text description feature matrix and the state association feature matrix through the graph neural network, as the state semantic association feature vector perpendicular to the sample dimension of the state semantic association feature matrix represents the state topology text description feature of a single electric device, the correlation of the state topology text description feature among a plurality of devices is possibly lower, so that the difficulty in adapting the parameters of the weight matrix of the classifier and the classification feature vector is increased in the training process of the classifier.

Based on this, in the technical solution of the present application, the parameters of the weight matrix of the classifier are adjusted in the training process, and the feature vector obtained after the state semantic association feature matrix is expanded is, for example, written asPerforming scene correlation optimization of classifier iteration, namely correcting the classification feature vector by using a weight matrix of the classifier before and after each iteration update according to the following formula to obtain the corrected classification featureA symptom vector; wherein, the formula is:

wherein the method comprises the steps ofRepresenting the classification feature vector,/- >And->Respectively representing the weight matrix before and after each iteration update of the classifier, and +.>Representing zero norm of vector,/->Representing addition by position->Representing difference by position +.>Representing matrix multiplication +.>An exponential operation representing a vector that represents a calculation of a natural exponential function value that is a power of a eigenvalue of each position in the vector.

In summary, the building integrated intelligent electricity utilization system 100 according to the embodiment of the present application is illustrated, which uses a first convolutional neural network model as a feature extractor to extract implicit association features of state information of all electric devices in a building to be monitored at a current time point, uses a context encoder including an embedded layer to extract global semantic association features described by type texts of each electric device in all electric devices, then uses a graph neural network to perform feature fusion, and finally uses a classifier to obtain a classification result for indicating whether the state combination of all electric devices in the building to be monitored is normal, and uses the classification result to generate abnormal electricity utilization prompt for the state combination of all electric devices in the building to be monitored. Therefore, the abnormality of the state combination of all the electric equipment in the building can be accurately analyzed and judged, and then an early warning prompt is generated for the abnormal situation.

As described above, the building integrated intelligent electricity usage system 100 according to the embodiment of the present application may be implemented in various terminal devices, such as a server for building integrated intelligent electricity usage, and the like. In one example, the building integrated intelligent electricity usage system 100 according to an embodiment of the present application may be integrated into the terminal device as one software module and/or hardware module. For example, the building integrated intelligent electricity system 100 may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the building integrated intelligent electricity system 100 may also be one of a plurality of hardware modules of the terminal device.

Alternatively, in another example, the building integrated intelligent electricity usage system 100 and the terminal device may be separate devices, and the building integrated intelligent electricity usage system 100 may be connected to the terminal device through a wired and/or wireless network and transmit the interactive information in a agreed data format.

Exemplary method

Fig. 5 illustrates a flow chart of a building integrated intelligent electricity usage method according to an embodiment of the present application. As shown in fig. 5, the building integrated intelligent electricity utilization method according to the embodiment of the application includes: s110, acquiring state information of all electric equipment in a building to be monitored at a current time point; s120, after state information of all electric equipment in the building to be monitored at the current time point is arranged into a state input vector, calculating the product between the state input vector and the transposed vector thereof to obtain a state association matrix; s130, the state association matrix is obtained through a first convolution neural network model which is completed through training and serves as a feature extractor; s140, obtaining the type text description of each electric equipment in all electric equipment; s150, respectively obtaining a plurality of device text description feature vectors through the trained context encoder comprising the embedded layer by the type text description of each electric device in all electric devices; s160, arranging the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension; s170, training the equipment text description feature matrix and the state association feature matrix through a trained graphic neural network to obtain a state semantic association feature matrix, wherein the graphic neural network encodes the equipment text description feature matrix and the state association feature matrix through learnable neural network parameters to obtain equipment text description association feature representation containing irregular state topology information; s180, training the state semantic association feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state combination of all electric equipment in a building to be monitored is normal or not; and S190, generating an abnormal electricity utilization prompt in response to the classification result that the state combination of all the electric equipment in the building to be monitored is abnormal.

Fig. 6 illustrates an architectural diagram of a building integrated intelligent electricity usage method according to an embodiment of the present application. As shown in fig. 6, firstly, in a network architecture of the building comprehensive intelligent electricity utilization method, acquiring state information of all electric equipment in a building to be monitored at a current time point; then, after the state information of all electric equipment in the building to be monitored at the current time point is arranged into a state input vector, calculating the product between the state input vector and the transposed vector thereof to obtain a state association matrix; then, the state association matrix is obtained through a first convolution neural network model which is completed by training and serves as a feature extractor; then, obtaining the type text description of each electric equipment in all electric equipment; then, the text description of the type of each electric equipment in all electric equipment is respectively passed through a context encoder which is completed by training and contains an embedded layer to obtain a plurality of equipment text description feature vectors; then, arranging the plurality of device text description feature vectors into a device text description feature matrix according to the device sample dimension; secondly, training the equipment text description feature matrix and the state association feature matrix through a trained graphic neural network to obtain a state semantic association feature matrix, wherein the graphic neural network encodes the equipment text description feature matrix and the state association feature matrix through learnable neural network parameters to obtain equipment text description association feature representation containing irregular state topology information; then, the state semantic association feature matrix is subjected to training through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state combination of all electric equipment in a building to be monitored is normal or not; and finally, generating an abnormal electricity utilization prompt in response to the classification result that the state combination of all electric equipment in the building to be monitored is abnormal.

In an embodiment of the present application, in the building integrated intelligent electricity utilization method, the training the state association matrix through a first convolutional neural network model as a feature extractor is further performed to obtain a state association feature matrix, and the method further includes: each layer of the trained first convolutional neural network model serving as the feature extractor is respectively carried out in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; carrying out mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the first convolutional neural network model serving as the feature extractor after training is the state association feature matrix, and the input of the first layer of the first convolutional neural network model serving as the feature extractor after training is the state association matrix.

In an embodiment of the present application, in the building integrated intelligent power consumption method, the training the context encoder including the embedded layer to obtain a plurality of device text description feature vectors includes: performing word segmentation processing on the type text description of each electric equipment in all electric equipment to convert the type text description of each electric equipment in all electric equipment into word sequences composed of a plurality of words; mapping each word in the word sequence to a word vector using an embedding layer of the context encoder to obtain a sequence of word vectors; and performing global-based context semantic coding on the sequence of word vectors using a translator of the context encoder to obtain the plurality of device text description feature vectors.

In an embodiment of the present application, in the building integrated intelligent electricity utilization method, the training the state semantic association feature matrix by using a classifier is performed to obtain a classification result, where the classification result is used to indicate whether a state combination of all electric devices in a building to be monitored is normal, and the method further includes: using the classifier to classify the state as followsProcessing the semantic association feature matrix to generate the classification result, wherein the formula is as follows:wherein->Representing projection of the state semantic association feature matrix as a vector,>weight matrix for full connection layer, +.>Representing the bias matrix of the fully connected layer.

In an embodiment of the present application, in the building integrated intelligent electricity usage method, the building integrated intelligent electricity usage method further includes: the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network, and the classifier are jointly trained.

In one embodiment of the present application, in the building integrated intelligent electricity utilization method, the training the first convolutional neural network model as the feature extractor, the context encoder including the embedded layer, the graph neural network and the classifier jointly includes: acquiring training data, wherein the training data comprises state information of all electric equipment in a building to be monitored at a preset time point and type text description of each electric equipment in the electric equipment; after state information of all electric equipment in a building to be monitored in the training data at a preset time point is arranged into a training state input vector, calculating the product between the training state input vector and a transpose vector thereof to obtain a training state association matrix; the training state association matrix passes through the first convolution neural network model serving as a feature extractor to obtain a training state association feature matrix; respectively passing the type text description of each electric equipment in all electric equipment in the training data through the context encoder comprising the embedded layer to obtain a plurality of training equipment text description feature vectors; arranging the training device text description feature vectors into a training device text description feature matrix according to the device sample dimension; the training equipment text description feature matrix and the training state association feature matrix pass through the graph neural network to obtain a training state semantic association feature matrix; expanding the training state semantic association feature matrix into classification feature vectors according to row vectors; correcting the classification feature vector by using a weight matrix before and after each iteration update of the classifier to obtain a corrected classification feature vector; passing the corrected classification feature vector through the classifier to obtain a classification loss function value; and jointly training the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network, and the classifier with the classification loss function value as a loss function value and by back propagation of gradient descent.

In an embodiment of the present application, in the building integrated intelligent power consumption method, the using the weight matrix of the classifier before and after each iteration update, correcting the classification feature vector to obtain a corrected classification feature vector, further includes: correcting the classification characteristic vector by using a weight matrix of the classifier before and after each iteration update according to the following formula to obtain the corrected classification characteristic vector; wherein, the formula is:

/>

Here, it will be understood by those skilled in the art that the specific functions and operations in the above-described building integrated intelligent electricity usage method have been described in detail in the above description of the building integrated intelligent electricity usage system with reference to fig. 1 to 4, and thus, repetitive descriptions thereof will be omitted.

Claims

1. An integrated intelligent electricity utilization system for a building, comprising:

a matrix construction unit, configured to arrange the plurality of device text description feature vectors into a device text description feature matrix according to a device sample dimension;

the power utilization control result generating unit is used for responding to the classification result that the state combination of all electric equipment in the building to be monitored is abnormal and generating a power utilization abnormality prompt;

the building comprehensive intelligent electricity utilization system further comprises a training module for performing joint training on the first convolutional neural network model serving as the feature extractor, the context encoder comprising an embedded layer, the graph neural network and the classifier;

The training module comprises:

the training data acquisition unit is used for acquiring training data, wherein the training data comprises state information of all electric equipment in the building to be monitored at a preset time point and type text description of each electric equipment in the electric equipment;

the training state association unit is used for arranging the state information of all electric equipment in the building to be monitored in the training data at a preset time point into a training state input vector, and calculating the product between the training state input vector and the transposed vector thereof to obtain a training state association matrix;

the training state association feature extraction unit is used for enabling the training state association matrix to pass through the first convolution neural network model serving as the feature extractor to obtain a training state association feature matrix;

the training equipment description semantic coding unit is used for respectively obtaining a plurality of training equipment text description feature vectors through the context encoder comprising the embedded layer by using the type text description of each electric equipment in all electric equipment in the training data;

the training matrix construction unit is used for arranging the training device text description feature vectors into a training device text description feature matrix according to the device sample dimension;

The training image neural network unit is used for enabling the training equipment text description feature matrix and the training state association feature matrix to pass through the image neural network so as to obtain a training state semantic association feature matrix;

the matrix unfolding unit is used for unfolding the training state semantic association feature matrix into classification feature vectors according to row vectors;

the classification vector correction unit is used for correcting the classification feature vector by using the weight matrix before and after each iteration update of the classifier so as to obtain a corrected classification feature vector;

the classification loss unit is used for passing the corrected classification characteristic vector through the classifier to obtain a classification loss function value; and

a parameter updating unit for performing joint training on the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network, and the classifier by using the classification loss function value as a loss function value and by back propagation of gradient descent;

the classification vector correction unit is further configured to: correcting the classification characteristic vector by using a weight matrix of the classifier before and after each iteration update according to the following formula to obtain the corrected classification characteristic vector;

Wherein, the formula is:

wherein V represents the classification feature vector, M ₁ And M ₂ Respectively representing weight matrixes of the classifier before and after each iteration update, wherein II is II ₀ Representing the zero norm of the vector, the zero represents the sum by position,representing difference by position +.>Representing matrix multiplication, exp (·) represents the exponential operation of the vector representing the computation of the natural exponential function value raised to a power by the eigenvalues of each position in the vector.

2. The building integrated intelligent electricity system according to claim 1, wherein the state association feature extraction unit is further configured to: each layer of the trained first convolutional neural network model serving as the feature extractor is respectively carried out in forward transfer of the layer:

carrying out convolution processing on input data to obtain a convolution characteristic diagram;

carrying out mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; and

non-linear activation is carried out on the pooled feature map so as to obtain an activated feature map;

the output of the last layer of the first convolutional neural network model serving as the feature extractor after training is the state association feature matrix, and the input of the first layer of the first convolutional neural network model serving as the feature extractor after training is the state association matrix.

3. The building integrated intelligent electricity system according to claim 2, wherein the device description semantic coding unit comprises:

the word segmentation subunit is used for carrying out word segmentation processing on the type text description of each electric equipment in all electric equipment so as to convert the type text description of each electric equipment in all electric equipment into a word sequence consisting of a plurality of words;

an embedded encoding subunit for mapping each word in the word sequence to a word vector using an embedding layer of the context encoder to obtain a sequence of word vectors; and

a context encoding subunit, configured to perform global-based context semantic encoding on the sequence of word vectors using a translator of the context encoder to obtain the plurality of device text description feature vectors.

4. The building integrated intelligent power consumption system according to claim 3, wherein the electric equipment state monitoring result generating unit is further configured to: processing the state semantic association feature matrix using the classifier to generate the classification result with the following formula: softmax { (M) _c ,B _c ) -Project (F) }, where Project (F) represents projecting the state semantic association feature matrix as a vector, M _c Weight matrix of full connection layer, B _c Representing the bias matrix of the fully connected layer.

5. The comprehensive intelligent electricity utilization method for the building is characterized by comprising the following steps of:

arranging the plurality of device text description feature vectors into a device text description feature matrix according to the dimension of the device sample;

generating an abnormal electricity utilization prompt in response to the classification result that the state combination of all electric equipment in the building to be monitored is abnormal;

the building comprehensive intelligent electricity utilization method further comprises the following steps: performing joint training on the first convolutional neural network model serving as a feature extractor, the context encoder comprising an embedded layer, the graph neural network and the classifier;

the jointly training the first convolutional neural network model as a feature extractor, the context encoder including an embedded layer, the graph neural network, and the classifier, comprises:

acquiring training data, wherein the training data comprises state information of all electric equipment in a building to be monitored at a preset time point and type text description of each electric equipment in the electric equipment;

after state information of all electric equipment in a building to be monitored in the training data at a preset time point is arranged into a training state input vector, calculating the product between the training state input vector and a transpose vector thereof to obtain a training state association matrix;

The training state association matrix passes through the first convolution neural network model serving as a feature extractor to obtain a training state association feature matrix;

respectively passing the type text description of each electric equipment in all electric equipment in the training data through the context encoder comprising the embedded layer to obtain a plurality of training equipment text description feature vectors;

arranging the training device text description feature vectors into a training device text description feature matrix according to the device sample dimension;

the training equipment text description feature matrix and the training state association feature matrix pass through the graph neural network to obtain a training state semantic association feature matrix;

expanding the training state semantic association feature matrix into classification feature vectors according to row vectors;

correcting the classification feature vector by using a weight matrix before and after each iteration update of the classifier to obtain a corrected classification feature vector;

passing the corrected classification feature vector through the classifier to obtain a classification loss function value; and

performing joint training on the first convolutional neural network model serving as a feature extractor, the context encoder containing an embedded layer, the graph neural network and the classifier by taking the classification loss function value as a loss function value and through back propagation of gradient descent;

Correcting the classification feature vector by the following formula to obtain a corrected classification feature vector;

wherein, the formula is:

6. The method for building integrated intelligent power consumption according to claim 5, wherein said training the state-related matrix through a first convolutional neural network model as a feature extractor to obtain a state-related feature matrix, further comprises: each layer of the trained first convolutional neural network model serving as the feature extractor is respectively carried out in forward transfer of the layer:

7. The method for comprehensive intelligent power utilization of building according to claim 6, wherein the step of obtaining a plurality of device text description feature vectors by training the context encoder including the embedded layer for each type text description of each of the all the electric devices includes:

performing word segmentation processing on the type text description of each electric equipment in all electric equipment to convert the type text description of each electric equipment in all electric equipment into word sequences composed of a plurality of words;

mapping each word in the word sequence to a word vector using an embedding layer of the context encoder to obtain a sequence of word vectors; and

the sequence of word vectors is globally based context semantic encoded using a translator of the context encoder to obtain the plurality of device text description feature vectors.